Installing crate breaks code

[dire spire]

Greetings. I'm a beginner rust developer.
I'm writing project to automate some of my web3 related stuff, so mostly I do browser automation.
For that purpouse I use chromiumoxide crate, with their fetcher feature.
The problem arise when I install ethers-rs crate, to be able to sign some user-interactions in automated manner.
Just installing this package makes my code panic with this error:

thread 'main' panicked at 'called `Result::unwrap()` on an `Err` value: BrowserFail(ChannelSendError(Send(SendError { kind: Disconnected })))'

It worth noting that there also tokio::task handler, which polls browser websocket from another thread and therefore emits additional panic:

thread 'tokio-runtime-worker' panicked at 'data did not match any variant of untagged enum Message', autoblur-node-rs\src\blur\browser.rs:74:21

Can't imagine how installing crate affects code in another crate (I exclude bug in my own code because without ethers-rs installation it works flawlessly)

[dire spire]

Cargo tree output: https://nekobin.com/cujimepudi

[peak cipher]

could be a case of features

https://doc.rust-lang.org/cargo/reference/features.html

if you and a dependency both depend on some crate, it will only be compiled once, and all the features will be combined

so maybe that dependency, somewhere in its dependency tree, enables a feature in a crate you use

that's not meant to happen, features are meant to be purely additive, but it can

[fleet tundra]

so the main thread panics first?

[dire spire]

No, the second thread panics first

[fleet tundra]

what does the code around like 74 look like?

[dire spire]

self.page = Some(
            self.browser
                .new_page("about:blank")   // <=== error here
                .await
                .map_err(move |error| Box::new(BlurInitError::BrowserFail(error)))
                .unwrap(),
        );

        let page = self.page.as_mut().unwrap();

        // load interceptor
        page.evaluate_on_new_document(REQ_INTERCEPT_SCRIPT.to_string().replace("{}", &pk))
            .await
            .map_err(move |error| Box::new(BlurInitError::BrowserFail(error)))?;

Mentioned tokio task handler which panics first

tokio::task::spawn(async move {
            while let Some(h) = handler.next().await {
                if h.is_err() {
                    panic!("{}", h.unwrap_err().to_string());
                }
            }
        }

Got handler as result from Browser::launch
https://docs.rs/chromiumoxide/latest/chromiumoxide/browser/struct.Browser.html#method.launch

[fleet tundra]

alright I think I found where the error is originating, but to get anything more I would need you to edit your local copy of the source for chromiumoxide

[dire spire]

Sure

[fleet tundra]

oh wait it has tracing built in

do you have tracing setup for the project by chance?

[dire spire]

Yes, I saw it too, so I have installed tracing as well. But because of my inexpirience I couldn't figure out how to turn it on

[fleet tundra]

do you have a subscriber?

If you don't I would recommend following https://docs.rs/tracing-subscriber/latest/tracing_subscriber/fmt/index.html#usage

what would be really helpful is to get the output of https://github.com/mattsse/chromiumoxide/blob/1eb7732171bec33223ca5dfd340e8afd8eb89e02/src/conn.rs#L145

hmm though its probably going to be the same error as above...

what I really want is the json

[fleet tundra]

of you could change line 139 (at least thats what it is on the repo) from

return match serde_json::from_slice::<Message<T>>(&msg.into_data()) {

to

return match serde_json::from_slice::<Message<T>>(dbg!(&msg.into_data())) {

I think that would be informative as then it should print out the json message the server sent

it will be in the form of a long list of numbers, but we can convert it to text if we get it

(https://github.com/mattsse/chromiumoxide/blob/1eb7732171bec33223ca5dfd340e8afd8eb89e02/src/conn.rs#L139C1-L139C86 for reference)

[dire spire]

https://nekobin.com/takivejuve

[fleet tundra]

On another note, after digging through your cargo tree output I think I might know what is happening

so it looks like chromiumoxide uses the async-std runtime by default

however either-rs polls in a package that switches things to use the tokio runtime

so maybe you need to set chromiumoxide to also use tokio

[dire spire]

Yes, I have already done it in my Cargo.toml

chromiumoxide = { git = "https://github.com/mattsse/chromiumoxide", features = ["tokio-runtime", "_fetcher-native-tokio"], default-features = false, branch = "main"}

[fleet tundra]

is this the features you used without ethers-rs?

[dire spire]

Yes

[fleet tundra]

well that idea goes out the window

it's either the browser sending back a "bad" message or the websocket implementation somehow

[dire spire]

How can I force cargo to recompile chromiumoxide ?

[fleet tundra]

you could do a cargo clean

though its usually good picking up changes by itself

[dire spire]

Here is output from a recompiled version:
https://nekobin.com/wuqapekugu

and last decoded output:

[fleet tundra]

Good response

{
  "id": 6,
  "result": {
    "frameTree": {
      "frame": {
        "id": "5D60618AD49142710F6B71FC4777CB0E",
        "loaderId": "DC528A83E5EE10D4FB59D26759340545",
        "url": "chrome://newtab/",
        "domainAndRegistry": "",
        "securityOrigin": "chrome://newtab",
        "mimeType": "text/html",
        "adFrameStatus": {
          "adFrameType": "none"
        },
        "secureContextType": "Secure",
        "crossOriginIsolatedContextType": "NotIsolated",
        "gatedAPIFeatures": []
      }
    }
  },
  "sessionId": "FEA4F65DBB9FB7573FEA7820079DA35B"
}

bad:

{
  "method": "Page.lifecycleEvent",
  "params": {
    "frameId": "5D60618AD49142710F6B71FC4777CB0E",
    "loaderId": "DC528A83E5EE10D4FB59D26759340545",
    "name": "commit",
    "timestamp": 106302.789599
  },
  "sessionId": "FEA4F65DBB9FB7573FEA7820079DA35B"
}

so um yeah that doesn't seem right

[dire spire]

hold on, I will remove ethers-rs and will provide you with a output

[fleet tundra]

k

I just decoded the one above the last to get that first one

so https://github.com/mattsse/chromiumoxide/blob/1eb7732171bec33223ca5dfd340e8afd8eb89e02/chromiumoxide_types/src/lib.rs#L205 is the type in question

it looks like an event maybe

/// A received `Event` from the websocket where the `params` is deserialized as
/// json
#[derive(Deserialize, Debug, PartialEq, Eq, Clone)]
pub struct CdpJsonEventMessage {
    /// Name of the method
    pub method: MethodId,
    /// The session this event is meant for.
    pub session_id: Option<String>,
    /// Json payload of the event
    pub params: serde_json::Value,
}

ah yeah

and wouldn't you know it session_id doesn't have a rename

though I need to check if the other code actually used the default

it doesn't

[dire spire]

without ethers

[fleet tundra]

Now I have no idea why you didn't get any events before

[dire spire]

Side note: it works

And what's more interesting is that it works with [email protected]

[fleet tundra]

alright go in and change https://github.com/mattsse/chromiumoxide/blob/1eb7732171bec33223ca5dfd340e8afd8eb89e02/chromiumoxide_types/src/lib.rs#L89 to have a serde rename to sessionId and see if it works with ethers-rs

[fleet tundra]

the ones here don't seem the have that field

so I guess the question is why does the browser respond with the session ID when ethers-rs is there

The serde rename was removed in commit https://github.com/mattsse/chromiumoxide/commit/3d9b6e6dede8ced81b07b69a5a656830526b0d76 apparently

[fleet tundra]

oh none of the out going have session ids

[dire spire]

still the same error with same ws message

[fleet tundra]

hmm you did #[serde(rename = "sessionId"] right?

[dire spire]

pub struct CdpJsonEventMessage {
    /// Name of the method
    pub method: MethodId,
    /// The session this event is meant for.
    #[serde(rename = "sessionId", skip_serializing_if = "Option::is_none")]
    pub session_id: Option<String>,
    /// Json payload of the event
    pub params: serde_json::Value,
}

[fleet tundra]

the serializing stuff isn't needed

because it only has Deserialize, but still that shouldn't change anything

[dire spire]

yeah that's what I thought

[fleet tundra]

I dont see what else would cause the deserialization to fail

[fleet tundra]

oh wait https://github.com/mattsse/chromiumoxide/blob/1eb7732171bec33223ca5dfd340e8afd8eb89e02/chromiumoxide_types/src/lib.rs#L190 doesn't have a session_id field at all

i wonder how that plays with untagged

[dire spire]

I wonder why EventMessage impl gets sessionId from params field

[fleet tundra]

yeah I noticed that to

[fleet tundra]

hmm both of these parse find in the playground

[dire spire]

I have added session_id field here, but still

also, cargo tree -d output
https://nekobin.com/cubilezogi

[fleet tundra]

I have no clue why it fails to deserialize that message